Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys
Abstract We have developed a deep-learning-based framework for understanding the individual and mutually combined contributions of different alloying elements and environmental conditions towards the pitting resistance of corrosion-resistant alloys. A fully connected deep neural network (DNN) was tr...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2022-08-01
|
| Series: | npj Materials Degradation |
| Online Access: | https://doi.org/10.1038/s41529-022-00281-x |
| _version_ | 1811188968817950720 |
|---|---|
| author | Kasturi Narasimha Sasidhar Nima Hamidi Siboni Jaber Rezaei Mianroodi Michael Rohwerder Jörg Neugebauer Dierk Raabe |
| author_facet | Kasturi Narasimha Sasidhar Nima Hamidi Siboni Jaber Rezaei Mianroodi Michael Rohwerder Jörg Neugebauer Dierk Raabe |
| author_sort | Kasturi Narasimha Sasidhar |
| collection | DOAJ |
| description | Abstract We have developed a deep-learning-based framework for understanding the individual and mutually combined contributions of different alloying elements and environmental conditions towards the pitting resistance of corrosion-resistant alloys. A fully connected deep neural network (DNN) was trained on previously published datasets on corrosion-relevant electrochemical metrics, to predict the pitting potential of an alloy, given the chemical composition and environmental conditions. Mean absolute error of 170 mV in the predicted pitting potential, with an R-square coefficient of 0.61 was obtained after training. The trained DNN model was used for multi-dimensional gradient descent optimization to search for conditions maximizing the pitting potential. Among environmental variables, chloride-ion concentration was universally found to be detrimental. Increasing the amounts of dissolved nitrogen/carbon was found to have the strongest beneficial influence in many alloys. Supersaturating transition metal high entropy alloys with large amounts of interstitial nitrogen/carbon has emerged as a possible direction for corrosion-resistant alloy design. |
| first_indexed | 2024-04-11T14:26:56Z |
| format | Article |
| id | doaj.art-cf0e68790f6f4dba93ce3f76e45066ee |
| institution | Directory Open Access Journal |
| issn | 2397-2106 |
| language | English |
| last_indexed | 2024-04-11T14:26:56Z |
| publishDate | 2022-08-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Materials Degradation |
| spelling | doaj.art-cf0e68790f6f4dba93ce3f76e45066ee2022-12-22T04:18:50ZengNature Portfolionpj Materials Degradation2397-21062022-08-016111010.1038/s41529-022-00281-xDeep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloysKasturi Narasimha Sasidhar0Nima Hamidi Siboni1Jaber Rezaei Mianroodi2Michael Rohwerder3Jörg Neugebauer4Dierk Raabe5Max-Planck-Institut für Eisenforschung GmbHMax-Planck-Institut für Eisenforschung GmbHMax-Planck-Institut für Eisenforschung GmbHMax-Planck-Institut für Eisenforschung GmbHMax-Planck-Institut für Eisenforschung GmbHMax-Planck-Institut für Eisenforschung GmbHAbstract We have developed a deep-learning-based framework for understanding the individual and mutually combined contributions of different alloying elements and environmental conditions towards the pitting resistance of corrosion-resistant alloys. A fully connected deep neural network (DNN) was trained on previously published datasets on corrosion-relevant electrochemical metrics, to predict the pitting potential of an alloy, given the chemical composition and environmental conditions. Mean absolute error of 170 mV in the predicted pitting potential, with an R-square coefficient of 0.61 was obtained after training. The trained DNN model was used for multi-dimensional gradient descent optimization to search for conditions maximizing the pitting potential. Among environmental variables, chloride-ion concentration was universally found to be detrimental. Increasing the amounts of dissolved nitrogen/carbon was found to have the strongest beneficial influence in many alloys. Supersaturating transition metal high entropy alloys with large amounts of interstitial nitrogen/carbon has emerged as a possible direction for corrosion-resistant alloy design.https://doi.org/10.1038/s41529-022-00281-x |
| spellingShingle | Kasturi Narasimha Sasidhar Nima Hamidi Siboni Jaber Rezaei Mianroodi Michael Rohwerder Jörg Neugebauer Dierk Raabe Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys npj Materials Degradation |
| title | Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| title_full | Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| title_fullStr | Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| title_full_unstemmed | Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| title_short | Deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| title_sort | deep learning framework for uncovering compositional and environmental contributions to pitting resistance in passivating alloys |
| url | https://doi.org/10.1038/s41529-022-00281-x |
| work_keys_str_mv | AT kasturinarasimhasasidhar deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys AT nimahamidisiboni deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys AT jaberrezaeimianroodi deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys AT michaelrohwerder deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys AT jorgneugebauer deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys AT dierkraabe deeplearningframeworkforuncoveringcompositionalandenvironmentalcontributionstopittingresistanceinpassivatingalloys |